> From: yyyyyyyyy@xxxxxxxxxxxx
> Date: Fri, 12 Jan 2001 18:58:53 +0100 (MET)
>
> >From: Paul Eggert <yyyyyy@xxxxxxxxxxx>
> >Under your scheme, I suppose we could rewrite the code like this:
>
> > sprintf (message,
> > #if HAVE_J_PRINTF_FORMAT_SPEC
> > _("Could only read %ju of %ju bytes"),
> > #else
> > _("Could only read %lu of %lu bytes"),
> > #endif
> > (uintmax_t) status, (uintmax_t) bytes);
>
> This will break on Solaris 8.
Yes, you're right of course. Sorry about that. Solaris 8 has a
longer-than-long uintmax_t, but it lacks %j.
This of course underlines the problems of the d5 approach. Not only
do the proposed transition strategies bloat application code: they are
hard to get right. Nobody has ever done this sort of conversion.
There may be other problems that we don't know about yet.
In other words, d5's proposed incompatible change to the standard
would force portable application writers into uncharted waters.
|